Greg Hudson <ghudson@MIT.EDU> writes:
> Is it me, or are we people implementing merging without a lot of design
> review on the dev list?
I think Greg's point is a good one.
Sorry for the rush, Greg. Ben, Mike, Greg Stein and I met, came up
with a plan, and just started coding. We need to write up a summary
of that plan and make sure it gets reviewed on the list, and of course
comments and constructive flames will be welcome. Will do as soon as
I get through the sixty remaining list emails here (what a week!)
In the scenario you describe below, the relatedness check, as we're
planning to use it, would succeed just fine. It wouldn't require you
to use `--force'.
Of course, you had no way of knowing that, because there's not yet any
clear description of what we're doing. Certainly our fault, not
yours.
-Karl
> On Thu, 2002-01-17 at 18:17, cmpilato@tigris.org wrote:
> > The svn_fs_check_related function compares two node IDs to see if they
> > are related, either through revision history, or through copy-from
> > history! This is the sanity check we need for 'svn merge' and 'svn
> > switch' (as in, "Um...these things you're trying to merge aren't even
> > related, so if you really want to do this, please pass --force to the
> > command.")
>
> I don't think this is a good idea.
>
> For instance, suppose I'm using "svn import" to bring in vendor versions
> of emacs into /emacs/20.1, /emacs/20.2, etc., and I have my own locally
> modified version in /emacs/trunk. When I want to update my local mods
> from version 20.1 to 20.2, I'll want to do a merge of the changes from
> /emacs/20.1 to /emacs/20.2 into /emacs/trunk. At least some of the
> pairs won't be related; do I really need to use --force for such a
> common scenario? It's bad to teach users to use --force when they're
> not doing something really odd, or they'll just use it all the time and
> lose all of the safeties.
>
> More philosophically, the "relatedness" check between nodes is suspect
> enough when it's only being used as a performance heuristic. Making it
> user-visible in ways like this seems like a great way to make Subversion
> appear complicated and baroque to users.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:57 2006